Research On The Prediction Method Of Protein-lncRNA Interaction Based On Deep Learning

Proteins play a major role in life activities.RNA is responsible for transmitting genetic information from DNA to proteins.However,the RNA that can encode proteins only accounts for about 2%,and the rest are non-coding RNA(nc RNA).nc RNA is closely related to the cell activity process and the pathological process of tumors,cancer,diabetes and other diseases.In addition,lnc RNA can only play its role when it is combined with protein.Therefore,the study of protein-lnc RNA interaction is of great significance for understanding the function of lnc RNA and protein.This article mainly uses the deep migration learning method to improve the index of protein-lnc RNA interaction prediction as the research goal.The main research contents include: Using the method of deep transfer learning based on lnc RNA sequence data,pre-train the RNA secondary structure prediction network E2 Efold,and use the encoder part of E2 Efold as the feature extraction network to obtain the feature expression of lnc RNA.Based on the protein sequence,using the method of deep migration learning,Alphafold2 is used as the encoder to predict the tertiary structure of the protein,and then a protein sequence design network is pre-trained,and the encoder of the pre-trained protein design network is used as the protein feature extraction To obtain the characteristic expression of the protein.Based on the characteristic expression of lnc RNA and protein,design a network to predict the interaction between lnc RNA and protein.Finally,the Deep RPI model was proposed,and leading results were obtained in the two benchmark datasets RPI369 and RPI488 for the prediction of the interaction between proteins with lnc RNA.Among them,the 5-fold cross-validated Acc,Sn,Sp,MCC,and AUC indicators on the RPI369 dataset are better than the comparison method;the 10-fold cross-validated ACC,MCC,and AUC indicators on the RPI488 dataset are all better method of comparison.